Professional Documents
Culture Documents
Professor and Marie Curie Chair in Fundamental and Conceptual Aspects of Turbulent Flows Institute for Mathematical Sciences and Department of Aeronautics, Imperial College London
Lectures series as a part of the activity within the frame of the Marie Curie Chair Fundamental and Conceptual Aspects of Turbulent Flows.
Arkady Tsinober
We absolutely must leave room for doubt or there is no progress and no learning. There is no learning without posing a question. And a question requires doubt...Now the freedom of doubt, which is absolutely essential for the development of science, was born from a struggle with constituted authorities... FEYNMANN, 1964
LECTURES V-VI
One of the important common features of processes resulting in turbulence is that all of them tend to enhance the rotational and dissipative properties of the flow in the process of transition to turbulence. The first property is associated with the production of vorticity (turbulence is highly rotational), whereas the second property is due to the production of strain (turbulence is strongly dissipative). The latter is a unique process of genuine turbulence and does not have any analogue in passive turbulence.
CHEN, 2000
One of the important common features of processes resulting in turbulence is that all of them tend to enhance the rotational and dissipative properties of the flow in the process of transition to turbulence. The first property is associated with the production of vorticity, whereas the second property is due to the production of strain.
- obsession
DISSIPATION/STRAIN Symmetric part of Aik = ui/xk, i.e. sik KOLMOGOROV 1941
- religion
1867
The turbulence syndrome includes the following symptoms:it is essentially nonlinear and rotational STEWART 1963 Turbulence is rotational and three dimensional. Turbulence is characterized by high levels of fluctuating vorticity. For this reason, vorticity dynamics plays an essential role in the description of turbulent flows. TENNEKES AND LUMLEY 1972 What is turbulence but a random chaotic field of vorticity, whose strong nonlinear interactions makes the problem so impossibly difficult? ... the concept of the coherent structure in turbulent shear flows has led to the picture of such flows as a superposition of organized, deterministic vortices whose evolution and interaction is the turbulence. SAFFMAN 1981
A great deal of interest is concentrated on the Reynolds stress tensor uiuj (or similar quantities in LES) and the closure problem, i.e. how it is related to the mean flow (with the not obvious - assumption that a relatively simple relation if at all - does exist). Such an approach is based on the view that small scales (whatever this means) are slaved to the large scales and are mostly a kind of passive sink of energy. This is the so called classical approach with the view that in order to understand turbulence one needs only to resolve the large scales and model the small scales (can this be done without sufficient understanding and/or full resolution down to all physically relevant scales?).
However, today it is high time to ask how much justified is such oversimplified treatment of small scales via methods like eddy viscosity, and similar. The small scales contain a great deal of essential physics of turbulent flows, much of which is not known or poorly understood, and which are intimately and bidirectionally related to the large scales (nonlocality). Apart of basic there are numerous problems in which one has to deal explicitly with the nature, structure and dynamics/evolution of small scales.
For instance, special information on small scale structure(s) is needed in problems concerning, e.g. combustion, disperse multiphase flow, mixing, cavitation, turbulent flows with chemical reactions, some environmental problems, generation and propagation of sound and light in turbulent environments, and some special problems in blood flow related to such phenomena as hemolysis and thrombosis. In such problems, not only special statistical properties are of importance like those describing the behaviour of smallest scales of turbulence, but also actual nonstatistical' features like maximal concentrations in such systems as an explosive gas which should be held below the ignition threshold, some species in chemical reactions, concentrations of a gas with strong dependence of its molecular weight on concentration (such as hydrogen fluoride used in various industries e.g. in production of unleaded petrol) and toxic gases.
Velocity derivatives, Aij = ui/xj, play an outstanding role in the dynamics of turbulence for a number of reasons. Their importance has become especially clear since TAYLOR, 1938 and KOLMOGOROV, 1941. Taylor emphasized the role of vorticity, i.e. the antisymmetric part of the velocity gradient tensor Aij = ui/xj, whereas Kolmogorov stressed the importance of dissipation, and thereby of strain, i.e. the symmetric part of the velocity gradient tensor. It is important (see below) that the whole (incompressible) flow field is fully determined by the fields of vorticity or strain with appropriate boundary conditions.
Apart from vorticity and strain/dissipation, there are many other reasons for special interest in the characteristics of the field of velocity derivatives, Aij=ui/xj, in turbulent flows:. In the Lagrangian description of fluid flow in a frame following a fluid particle, each point is a critical one, i.e. the direction of velocity is not determined. So everything happening in its proximity is characterized by the velocity gradient Aij=ui/xj. For instance, local geometry/topology is naturally described in terms of critical points terminology. The field of velocity derivatives is much more sensitive to the non-Gaussian nature of turbulence or more generally to its structure, and hence reflects more of its physics. The possibility of singularities being generated by the Euler and the Navier-Stokes equations (NSE) and possible breakdown of NSE are intimately related to the field of velocity derivatives.
There is a generic ambiguity in defining the meaning of the term small scales (or more generally scales) and consequently the meaning of the term cascade in turbulence research. The specific meaning of this term and associated interscale energy exchange/`cascade' (e.g. spectral energy transfer) is essentially decomposition/representation dependent. Perhaps, the only common element in all decompositions/representations (D/R) is that the small scales are associated with the field of velocity derivatives. Therefore, it is natural to look at this field as the one objectively (i.e. D/R independent) representing the small scales (one can following CORRSINuse higher order derivatives, e.g. curl , sij ). Indeed, the dissipation is associated precisely with the strain field, sij, both in Newtonian and nonNewtonian fluids. There is a number of more specific reasons why studying the field of velocity derivatives is so important in the dynamics of turbulence.
Vorticity and strain are not just velocity derivatives. They are special for several reasons as mentioned (and will be discussed at length below). The one to be stressed here is that the whole flow field is determined entirely by the field of vorticity or strain *with appropriate boundary conditions: u = -curl ; ui=2sij/xj, i.e. the velocity field is a linear functional of vorticity u = F{}, or strain ui = G{sij},*i.e. alteration of the field of velocity derivatives reflects on the velocity field, vorticity and strain are not passive they react back and not only for the above (kinematic reason)
In the whole space the first functional is the well known Bio-Savart law, and the second has a similar form (see eq. C.14 in Tsinober 2001) jj (r)= - (4)-1 ijk rj/r, ri = xi-yi, js(r)= - (2)-1 rj/r, ri = xi-yi, ij (r)= - 3(8)-1 {ijlrlrk+ kjlrlri}/r5, ri = xi-yi, P.V. stands for the Cauchy principal value.
Strain controls the flow in the same way as does vorticity also in the sense of possible breakdown of smooth solutions for 3D flows, Ponce, G. (1985) Remarks on a paper J.T. Beale, T.
Kato and A. Majda, Commun. Math. Phys., 98, 345-353.
Dolphins in phosphorescent sea. The inspiration for this woodcut, created by M. C. Escher in 1923, was the flow-induced bioluminescence that occurs on dolphins when they swim through waters that contain high levels of bioluminescent plankton.
Some reasons
More below
flows are defined by strain 2. Energy dissipation is directly associated with strain (both Newtonian and non-Newtonian) and not with vorticity. 3. The energy cascade (whatever this means) and its final result dissipation, are associated with predominant self-amplification of the rate of strain and vortex compression rather than with vortex stretching. That is another nonzero odd moment - sijsjksjk responsible for the production of strain, is not less important than the enstrophy production ijsij
4. Vortex stretching is essentially a process of interaction of vorticity and strain. Vortices interact via their strain fields. 5. Strain dominated regions appear to be the most active/nonlinear in a number of aspects. 6. Interaction of the flow field with additives (particles, polymers, blood cells) is mostly via strain . 7. Though formally all the flow field is determined entirely by the field of vorticity the relation between the strain and vorticity is strongly nonlocal. In many cases, they are only weakly statistically correlated or not correlated at all.
BASIC EQUATIONS
D/Dt = ()u + + ijkFk/xj
()D/Dt = ijsij + ii + ijki Fk/xj
Dsij/Dt = - sjkski (1/4)(ij- 2ij) p/xixj + sij + Fij ()Ds/Dt = - sijsjkski (1/4)ijsij sijp/xixj + sijsij + sijFij
In what sense vortex stretching plays a central role in the energy cascade to small scales and dissipation? OR
One of the most basic phenomena and distinctive features of three-dimensional turbulence is the predominant vortex stretching. This process occurs via interaction of vorticity and strain.
It is a very common view that this process is responsible for the enhanced dissiption in 3-D turbulent flows
It seems that the stretching of vortex filaments must be regarded as the principal mechanical cause of the high rate of dissipation which is associated with turbulent motion, TAYLOR 1938
Vortex-line stretching plays a central role in the energy cascade to small scales and dissipation. RHINES, 1997 (p.102)
2 2s =
BUT, dissipation is
2sijsij !!
The true physical causal relation is between dissipation and strain both in Newtonian and nonNewtonian fluids. Therefore it is a misconception to associate dissipation directly with vorticity. However, the rotational nature of turbulence (i.e. vorticity) is crucial for dissipation.
As a home work show that in irrotational flow (i.e. u=) the dissipation is either zero or volume integral 2s2dV over the flow domain is equal to the surface integral u2dV
There exist two nonlocally connected and weakly correlated processes. The second is the self-amplification of strain.
NONLOCALITY OF
VORTICITY/STRAIN RELATION AND THE
ISUE OF SURROGATES
Field experiment 2000, Israeli field station, Re=104
THE PROBE
3 mm
hot wires
cold wires
Manganin is used as a material for the sensor prongs instead of tungsten because the temperature coefficient of the electrical resistance of manganin is 400 times smaller than that of tungsten.
SELFAMPLIFICATION
SELFAMPLIFICATION
SELF-RANDOMIZATION/INTRINSIC STOCHASTICITY: NO SOURCE OF RANDOMNESS IS NEEDED, THE FORCING CAN BE CONSTANT IN TIME
AT THE LEVEL OF VELOCITY DERIVATIVES: VORTICITY AND STRAIN (DISSIPATION) THE EXTERNAL FORCING IS IRRELEVANT
Three cases: 1. DNS in a periodic box, Re=102 2. DNS in a channel flow, Re=5600 3. Atmospheric SL, Re=104; Re=108
0.5 F O R C I N G
5 0
Assume there is no production of enstrophy in the mean ijsij = 0. Is there turbulence? What about a similar assumption for strain production sijsjkski = 0 ?
.
Enstrophy production is approximately balanced by viscous terms at any - whatever large - Reynolds number?*
TENNEKES AND LUMLEY (1972, P.91):
Note i) approximate balance between P and --D and ii) irrelevance of the forcing term F
ARE VORTEX LINES (OR VORTICITY) APPROXIMATELY FROZEN IN FLUID FLOW AT LARGE REYNOLDS NUMBERS?
...
a material line which is initially coinsides with a vortex line continues to do so. It is thus possible and convenient to regard a vortexline as having a continuing identity and as moving with the fluid (In a viscous fluid it is, of course, possible to draw the pattern of vortex lines at any instant, but there is no way in which particular vortex-line can be identified at diffrent instants). BATCHELOR, 1967, p. 274
ijsij =
2{k
cos2(
,k)}
BETCHOV 1956
k - eigenvectors of the rate of strain tensor sij, 1 > 2 >3; 1+2 +3 = 0 (div u=0); 1 > 0, 3<0;
2{1 cos2(,1)} = I 2{2cos2(,2)} = II 2{3 cos2(
I: II: III = 3: 1: -1
,3)} = III
EIGEN CONTRIBUTIONS
Value
2 cos 2 (, )
1 2 3 1 2 3 1 2 3 1 2 3
0.8m 1.44 0.44 -0.87 0.53 0.09 0.38 1.77 0.47 -1.24 0.51 0.08 0.41
1.2m 1.60 0.62 -1.22 0.33 0.05 0.63 1.56 0.50 -1.07 0.50 0.09 0.41
2.0m 1.36 0.67 -1.03 0.29 0.15 0.56 1.63 0.52 -1.15 0.50 0.10 0.41
3.0m 1.31 0.46 -0.77 0.52 0.14 0.34 1.91 0.45 -1.36 0.51 0.10 0.40
4.5m 1.53 0.46 -0.99 0.46 0.13 0.41 2.08 0.47 -1.55 0.49 0.10 0.41
7.0m 1.04 0.58 0.49 0.16 0.35 1.55 0.54 0.50 0.11 0.40
10m 1.37 0.49 0.49 0.15 0.36 2.19 0.47 0.49 0.10 0.41
-0.62 -0.85
2 2 2 cos (, )
cos 2 (, )
-1.09 -1.66
2 (, ) 2 cos
I II
Note that the dominating term is the first one (I) associated with the first eigenvector 1 corresponding to the largest (purely positive) eigenvalue. This is Yet vorticity is preferentially aligned with the second eigenvector 2 which is quite a bit counterintuitive.
ALIGNMENT BETWEEN THE EIGENFRAME I OF THE RATE OF STRAIN TENSOR Sij AND VORTICITY
10m 0.47 0.06 -0.53 0.41 0.06 0.55 0.60 0.02 -1.04
1 2 3
-0.62 -0.60
2 2 / s
1 2 3
2 3 / s
3/2
1 2 3
-0.73 -0.83
So what about - sijsjkski > 0 (in homogeneous flows equal to - 4/3 ijsij ) is it due to predominant stretching? sijsjkski = 13 +23 +33 = 312 3
positive due to dominant contribution from - 33 , i.e. predominant compressing rather than stretching !
There are several other (see the examples below) important processes in which predominant compressing is the main player : there is no reason to push everywhere stretching, especially vortex stetching.
MORE EXAMPLES WITH PREDOMINANT CONTRIBUTION OF COMPRESSING TKE PRODUCTION IN TURBULENT SHEAR FLOWS uiukSik
EVOLUTION OF DISTURBANCES IN
GENUINE AND PASSIVE TURBULENCE
PRODUCTION OF VORTICITY
DIMENSIONAL TURBULENCE
GRADIENTS IN TWO-
The turbulent energy production in a turbulent shear flow is known to be represented by the term -<uiuk>Sik, with ui being the components of velocity fluctuations, and Sik the mean rate of strain. In turbulent flows which are two-dimensional in the mean (i.e. such that /x3 = 0) the production term can be represented as where u = u1 + u2, iS are the eigenvalues and iS are the corresponding eigenvectors of the mean rate of strain tensor Sik, and 1S>0,2S<0 (3S=0). Since the term associated with the stretching of material elements is negative, - u1Scos(u,1S) < 0, and the term associated with the compressing of material elements is positive, u2Scos(u,1S) > 0, the production term -<uiuk>Sik can be (and usually is) positive due to positiveness of the term associated with the compressive (negative) eigenvalue/eigenvector {2, 2S}, of the mean strain Sik. In this sense the turbulent energy production is due to the predominant compressing of material elements rather than stretching.
COMPRESSING? Turbulent channel flow (same in BL) Phys Fluids, 16, 2704 (2004)
The main feature - tendency of alignment of the vector u with both 1S and 2 S. The difference is that the latter alignment is somewhat stronger, which results in the positive value of the TKE production
TKE PRODUCTION:
STRETCHING OR COMPRESSING?
STRETCHING OR COMPRESSING?
TKE PRODUCTION:
GROWTH OF DISTURBANCES IN GENUINE (Eu, E, Es) AND PASSIVE (EA , EB , E , EG) TURBULENCE
Looking at the evolution of the disturbance u of some flow realization u in a statistically steady state and similarly for other quantities.
For more details see Tsinober and Galanti 2003, Phys. Fluids, 15, 3514-3531.
uiujsij
Tsinober & Galanti, 2003
PRODUCTION OF VORTICITY
(1/2)D2/Dt = - ik sik + ii
The main contribution to the production (-iksik > 0) - iksik = - 2kcos2(,k) is associated with compressive eigenvalue 3: it is due to -23 cos2(; 3). Note that partner to i has the same partner as does vorticity (strain), but they (i and sik) are not equal partners as i lives at much smaller scales than sij and there is no strain production (!) either. However, production of palinstrophy, i.e. (curl )2 is due to predominant contribution from the term associated with the stretching eigenvalue 1.
PRODUCTION OF GRADIENTS/ DISSIPATION OF PASSIVE SCALAR D/Dt = - ()u + ; G = (1/2)DG (1/2) 2/Dt = - GiGk sik + GiGi
The main contribution to the production (-GiGksik > 0)
- GiGksik
= - G2kcos2(G;k)
is associated with the alignment of the temperature gradient and the eigenvector 3 corresponding to the compressive eigenvalue 3: it is due to -G23 cos2(G; 3).
Selected results
The PDF of -GiGjsij is positively skewed and the mean <-GiGjsij > is positive.
1D 2 2 G = Gi G j sij + Gi Gi 2 Dt
Alignments between the temperature gradient G and the eigenframe i of the rate of strain tensor sij
(k are the corresponding eigenvalues); 1> 2> 3; 1> 0, 3<0; 1+ 2+ 3 = 0
The main effect, the alignment between the temperature gradient and the eigenvector 3 corresponding to the compressive eigenvalue 3< 0, is captured well in the measurements and is similar to that obtained from DNS
JOINT PDFS OF ENSTROPY (top) AND STRAIN (bottom) PRODUCTION WITH ENSTROPHY (left) AND STRAIN (right)
FROM OUR FIELD EXPERIMENT AT Re=104 Physics of Fluids,
13, 311 (2001)
( / )
2 2 i i
( s
i 2 i
i 2 i / 2
k ik
/ 2
2
( / )
i
i k sik / 2
i 2 i / 2
Rate of enstrophy production and its viscous reduction conditioned on strain and vorticity
NOTE THE LARGE RATE OF ENSTROPHY PRODUCTION IN STRAIN DOMINATED REGIONS (RED CURVE) AS COMPARED TO REGIONS OF LARGE ENSTROPHY (BLUE CURVE)
( s
i
k ik
/ 2
)
2
i k sik /
s / s ,/
Eigen-contributions to ijsij/2=kcos2(,k)
Non-linear terms are growing much slower in the enstrophy dominated regions than in the strain dominated regions.
Summary of threedimensional, incompressible flow patterns/ local structure of the M RR flow field in the frame following a fluid particle
(from Soria et al. 1994, after Perry et al. 1990).
R-Q - PLOT
Field experiment, Re= 6800 PTV, Re= 80 DNS of NSE, Re= 690
velocity gradient
Q=
1 2 2 sik sik - second invariant of the velocity gradient tensor 4 1 3 R = sik skm smi + i k sik - third invariant of the velocity gradient tensor 3 4
The first invariant is vanishing as a consequence of incompressibility
Chacin et al 2000
ui xk
D=0
Q Q
D=0
R
Ejections (-u+v)
R
D=0
/(u4)
0.36
One of the most interesting (from our point of view) findings is that the main contribution to the shear stress, turbulent energy production, and dissipation comes from the regions with Q<0 with larger contribution from the lower right quadrant, i.e. Q<0 and R>0, not only dominated by strain, but also by production of strain, -sikskmsmi, see figure 5, also figures 8, 11 and 14 in Chacin and Cantwell (2000). It should be emphasized that these regions are mainly not
the ones corresponding to vortices (hairpins or whatever), which are located mostly in the regions with Q>0, or a bit more precisely in regions with D>0, where D=((27)/4)Q+R is the discriminant of ui/xj. That is the regions of major nonlinear activity are really associated with large strain (mainly corresponding to what Chacin and Cantwell call `blank' spaces) rather than with regions of concentrated vorticity with lower dissipation.
In other words it seems that concentrated vorticity is not that important also in turbulent shear flows and that structure(s) associated with turbulence (not only its energy) production are mainly due to the large strain rather than large vorticity. Structure(s) associated with the latter seem to be the consequence of the turbulent dynamics rather than its dominating factor. A final remark is that these results do not contradict the importance of vorticity in maintaining the Reylolds stress. First, these are relations for the mean quantities, and second, there is no turbulent flow without vorticity. However, important details of the relations between Reynolds stress, vorticity, strain and their production remain not clear.
The interpretation of the results by Chacin and Cantwell, 2000 (and similar) given here is not in full agreement with their conclusions, especially regarding the role of vortices and concentrated vorticity in turbulent flows (again vortex obsession).
At this stage, this alternative approach (i.e. the structural) has not led to a generally applicable quantitative model, neither for better or worse has it a major impact on the statistical approaches. Consequently the deterministic viewpoint is neither emphasized nor systematically presented, POPE 2000. This does not mean that there exists generally applicable quantitative model based on statistical approaches.
REAL
GAUSSIAN
(a)Big structure with ||>3; (b)Small structure with ||>3; (c)Big structure with ||>6
SIMILAR TO THOSE IN SHE et al., 1991 and BORATAV AND PELZ, 1997 CHEN, 2000
b
(a) |s|>2.8s; (b) |s|>4.2s
THE AMOUNT OF COMPRESSING IN WORMS IS THE SAME AS IN THE WHOLE FIELD ! HENCE INADEQUATE REPRESENTATION OF THE FLOW FILED BY A COLLECTION OF PURELY STRETCHED VORTICES (or other `simple' objects), ESPECIALLY THOSE
DESCRIBED BY DIFFRENTIAL EQUATIONS. THE REAL ONES ARE DESCRIBED BY INTEGRO DIFFERENTIAL EQUATIONS.
SOME COMMENTS ON
WHAT IS A VORTEX ?
I DO NOT KNOW. WHO DOES? THERE ARE SOME DEFINTIONS (MOSTLY AD HOC). A COMMON FEATURE IS THAT (MORE OR LESS) CONCENTRATED VORTICITY IS SURROUNDED BY PLENTY OF STRAIN , E. G. POTENTIAL VORTEX: VORTICITY ONLY IS NOT REALLY ONLY
A popular method to look for structure(s) (and `vortices) is to use a criterion based on one parameter only, e.g. enstrophy . Though such an approach is useful and `easy', it is inherently limited and reflects the simplest aspects of the problem. For example, even for characterization of some aspects of the local (i.e. in a sense `point'-wise) structure of the flow field in the frame following a fluid particle requires at least two parameters: the second and the third invariants of the velocity gradient tensor ui/xk: Q = 1/4( - 2sijsij) and R = - 1/3(sijsjkski + 3/4ijsij). Therefore attempts to adequately identify/characterize finite size structure(s) and vortices seem to be of this kind - by one parameter only are unlikely to be successful, and one needs something like pattern recognition based on some conditional sampling scheme involving definitely more (perhaps much more) than two parameters. But then all the simplicity (and attraction) will be gone.
So it is safer to use well defined quantities vorticity and strain. Vorticity alone is not enough. Concentrated vorticity (very popular) is not that important and other regions (e.g. regions dominted by strain) play essential role in the evolution and dynamics of turbulent flows (Tsinober, 1998, 2001)
IN LIEU OF CONCLUSION
Looking in more essential details (both physics and the mathematical ones) of the processes of self- amplificiation of the field of velocity derivatives seems to be the key issue for most of the problems of turbulence but also 3-D NSE and Euler. This includes subtle geometrical relations between vorticity and strain (which are likely to be the main guilty of almost happening in turbulence: after all there is no turbulence without vorticity production) and several others of dynamical significance . Understanding of these processes and thereby essential aspects of turbulence physics seems to form the basis for constructive approach to a great variety of problems including effective handling of nonlinearities which will allow to solve the standard 3D-NSE and Euler problems and not the other way round.